Every setting in Calls → Settings: recording, hold music, announcements, routing, the OpenAI key, the Voice AI engine, WhatsApp calling per number, and phone (PSTN) calling over SIP.
Where to find itSidebarCallsConfigurationSettingsOpen in Libromi
Before you start
The Call Center must be switched on for your workspace; phone (PSTN) calling is a beta that support enables.
Call Settings holds everything that controls voice calls in your workspace: WhatsApp calls, live chat calls and phone calls. This page lists each card and what it changes. For taking and making calls, see WhatsApp calling and the Calls center.
Where to find it
Sidebar → Calls → Configuration → Settings. Three tabs: Workspace, WhatsApp Calling and SIP / Phone (Beta).
Only workspace managers ("manage settings") can change settings. Others see: "These settings are managed by a workspace manager. You can view them below."
Before you start
Your plan must include WhatsApp voice calls (whatsapp_voice_call) and the Call Center must be active for your workspace. See WhatsApp calling and the Calls center.
Workspace tab
Card
Settings
Recording & transcription
Record calls, Transcribe calls, AI summary, Retention (days) ("Auto-delete recordings after N days (blank = keep)")
Music on hold
Play hold audio; upload an MP3 (15 MB at most) or reset to the default
Announcements
Recording consent announcement ("Played to the caller before the agent connects."), Voicemail greeting
Call routing
Mode (AI, Human, Hybrid), Default AI agent, No-agent fallback (AI, Queue, Voicemail, Callback), Human-direct, Allow AI after hours, Default queue, Assigned-staff-first, Routing strategy (Round-robin, Least-loaded), Max calls per agent; shown but not editable: Ring timeout (seconds) (system default) and Max concurrent calls (managed by operations)
OpenAI API key
Your own OpenAI key, "Used for spoken announcements, transcription and call summaries." Add key / Replace key. The key is checked with OpenAI before saving and never shown again
Voice AI engine
See below
Applies to on the recording and routing cards lets you override them for one number or one website.
Voice AI engine
"Which OpenAI engine answers your voice calls — website, WhatsApp and phone calls alike."
Setting
What it does
Standard
"Takes turns speaking and listening. Works with the shared key or your own."
Live (Beta)
"Talks and listens at the same time, so interruptions feel natural." Billed per minute on your own OpenAI account (the panel quotes about $0.05 per minute plus a small amount for the backend model)
Model, Voice
The model and voice for the chosen engine
Backend model (Live only)
"Luna is the low-cost brain; Terra reasons better and costs more."
Backend instructions (Live only)
Extra instructions for the backend model
Go to Calls → Settings, Workspace tab, Voice AI engine card
For Live, add your key first in the OpenAI API key card
Choose Standard or Live, then a Model and a Voice
Click Save
For Live, the panel checks with OpenAI that Live works on your account before saving. Changes apply to new calls right away.
WhatsApp Calling tab
Per number: WhatsApp number, Enable calling, Call button (Visible or Hidden for all), Auto-grant callback permission, Limit to business hours with a time zone and hours ("Outside these hours the call button is hidden."). SIP (advanced) → Enable SIP, Hostname, Port forwards that number's calls to your own SIP server: "Leave off unless you run a PBX." Settings sync with the WhatsApp Cloud API.
SIP / Phone tab (Beta)
Phone (PSTN) calling: "Make + receive calls on real phone numbers through your SIP provider — same AI agent, human handoff, recording and multi-line UI as WhatsApp calls."
Part
What it does
Enable PSTN for this workspace
Turns phone calling on or off. Needs a trunk first
SIP trunks → Add trunk
Provider, Name (optional), Outbound host, Numbers (DIDs, comma-sep) in E.164 (for example +15555550100), Auth username, Auth password ("Stored securely, never shown again"), Allowed signaling IPs (comma-sep), then Provision
Phone numbers (DIDs)
Numbers attached to a trunk. Add a DID and pick the trunk
Before phone calls connect
A checklist: platform PSTN switch on (managed by support), a trunk, at least one number, PSTN enabled
Limits and plans
Calling needs whatsapp_voice_call. Live chat calls need the web_voice add-on.
Phone (PSTN) calling is in beta. If it is not active you see: "Phone (PSTN) calling is in beta and isn't available for your workspace yet. Contact your support manager to request access."
Live engine minutes are billed on your own OpenAI account.
Good to know
Without your own key, Standard runs on the shared key; Live cannot be chosen: "Add your OpenAI key above first — Live runs on your own account, billed per minute."
If your saved model is no longer offered, pick another one before you can save.
Troubleshooting
What you see
Why
What to do
"These settings are managed by a workspace manager…"
You cannot manage settings
Ask a manager
Live greyed out, "Live isn't available on this workspace yet."
Live is not offered for your workspace
Use Standard, or contact support
"This model isn't offered any more. Pick another one to save changes."
The saved model was retired
Pick a listed model
"File too large" / "Max 15 MB."
Audio file over 15 MB
Upload a smaller MP3
Enable PSTN for this workspace greyed out
No trunk yet
Add a trunk first
"“…” isn't valid E.164."
A number is not in +countrycode format
Write it like +15555550100
"Provisioning failed"
The provider details were refused
Check host, credentials and allowed IPs with your provider